|Home | About | Journals | Submit | Contact Us | Français|
BLM, one of the human RecQ helicases, plays a fundamental role in homologous recombination-based error-free DNA repair pathways, which require its translocation and DNA unwinding activities. Although translocation is essential in vivo during DNA repair processes and it provides a framework for more complex activities of helicases, including strand separation and nucleoprotein displacement, its mechanism has not been resolved for any human DNA helicase. Here, we present a quantitative model for the translocation of a monomeric form of BLM along ssDNA. We show that BLM performs translocation at a low adenosine triphosphate (ATP) coupling ratio (1 ATP consumed/1 nucleotide traveled) and moderate processivity (with a mean number of 50 nucleotides traveled in a single run). We also show that the rate-limiting step of the translocation cycle is a transition between two ADP-bound enzyme states. Via opening of the helicase core, this structural change may drive the stepping of BLM along the DNA track by a directed inchworm mechanism. The data also support the conclusion that BLM performs double-stranded DNA unwinding by fully active duplex destabilization.
DNA and RNA helicases are nucleotide triphosphate (NTP)-consuming motor enzymes generating single-stranded (ss) forms of nucleic acids during fundamental cellular processes including replication, recombination and genome repair. RecQ-family DNA helicases are members of helicase superfamily 2 (SF2) involved in homologous recombination (HR)-based error-free DNA repair of double-stranded (ds) DNA breaks (DSBs) (1–4). Loss of function of BLM, one of the human RecQ helicases, causes Bloom’s syndrome, a severe autosomal genetic disease leading to high cancer predisposition (5,6). BLM was recently proposed to suppress or promote HR depending on cellular conditions (7). BLM exerts its anti-recombinogenic activity by preventing premature initiation of HR via dissociation of Rad51 nucleoprotein filaments. This anti-recombination role of BLM is reflected in the hyperrecombination phenotype of Bloom’s syndrome cells (5,8). On the other hand, BLM has also been proposed to be a key player in the synthesis-dependent strand annealing (SDSA) pathway of HR-based DSB repair (9). BLM can dissolve double Holliday junctions (10,11) and D-loops (12,13), and also stimulates DNA synthesis on model replication forks resembling one end of the D-loop by unwinding the displaced strand (7). In these processes, BLM must be able to perform adenosine triphosphate (ATP)-driven translocase, DNA helicase and branch migration activities.
Despite the fact that the biological functions of RecQ helicases have been widely investigated (14), their mechanisms of action are poorly understood. ATP-dependent translocation along ssDNA is the basis for the separation of the strands of dsDNA. Mechanistic knowledge about translocation thus provides a framework for more complex activities of DNA helicases, including dsDNA unwinding, branch migration and dissolution of double Holliday junctions (15). Moreover, ssDNA translocation itself is likely essential in vivo in DNA repair. At the initial phase of HR-based DSB repair, the BLM-induced disruption of Rad51 nucleoprotein filaments requires ssDNA translocation activity of BLM (7). Similarly, the translocation activity of the yeast Srs2 helicase is necessary for the catalysis of Rad51 nucleoprotein filament disassembly (16). ssDNA translocation-based unwinding of model replication forks also stimulates DNA synthesis by η polymerase, which role was reported in DSB repair (17), supporting a proposed key function of BLM in SDSA (9). SDSA also involves facilitated annealing of complementary ssDNA stretches that form after D-loop disruption (2,18). The capability of translocation along ssDNA is likely essential in vivo for efficient strand annealing.
In biochemical and genetic studies and crystal structures of UvrD, there remains a controversy whether the monomeric form of this enzyme is capable of unwinding dsDNA (19–23). Studies on PcrA and Rep demonstrated that these SF1 helicases are able to perform translocation but are unable to unwind dsDNA substrates in their monomeric forms (24,25). By contrast, N-terminal truncation of BLM abolishes oligomerization, while the monomeric enzyme form retains its ATPase and DNA unwinding activities (26,27). These findings suggest a different mode of coupling between translocation and strand separation activities of BLM from those of the bacterial SF1 helicases.
Translocation along ssDNA requires processive movement of the helicase: the motor must be able to perform multiple enzymatic cycles and coupled mechanical steps before dissociating from its track. To date, three different models have been proposed to explain the translocation mechanisms of SF1 and SF2 helicases: (i) The ‘inchworm’ stepping mechanism for the monomer form of PcrA (28) and UvrD (19); (ii) the ‘Brownian motor’ (thermal ratchet) for hepatitis C virus NS3 helicase (29); and (iii) a ‘non-uniform stepping model’ for UvrD (30) and RecBCD DNA helicases (31) and NS3 helicase (32). The inchworm and Brownian models assume the hydrolysis of a single ATP molecule per kinetic step, dictated by the rate-limiting step of the cycle, which also defines the DNA-activated steady-state ATPase activity. The inchworm mechanism supports unidirectional processive stepping, while a Brownian motor oscillates between weakly and tightly DNA-bound states with diffusion-driven movements occurring in the weakly bound states. Therefore, enzymes using the inchworm stepping mechanism perform processive translocation at a low ATP coupling ratio [≤1 ATP hydrolyzed/nucleotide (nt) traveled], while those using the Brownian ratchet model generally translocate at a higher ATP coupling ratio.
In this study, we determined all key parameters of the ATP-driven translocation of BLM along ssDNA, using signals directly monitoring ATP consumption as well as the interaction of the enzyme with ATP and DNA at high temporal resolution (Figure 1, Table 1). We demonstrate that the monomeric form of BLM is a moderately processive DNA translocase (on average, performing 50 ATPase cycles during a processive run) with a low ATP coupling ratio (1 ATP consumed/1 nt traveled). During translocation, neither the ATP hydrolysis step nor any product release steps are rate limiting in the kinetic cycle. We propose that BLM translocates along ssDNA using an active, probably inchworm-like mechanism in which the rate-limiting step is a structural transition between two ADP-bound states, which may directly lead to stepping along ssDNA. Our results also show that the rate of ssDNA translocation matches that of dsDNA unwinding (27), which supports a model in which BLM actively destabilizes the DNA duplex to perform rapid and efficient strand separation.
Cloning and purification of BLMHM as well as preparation of materials are described in Supplementary Data. All measurements were done at 25°C. DNA concentrations are expressed as those of oligo- or polynucleotide molecules (as opposed to those of constituent nucleotides).
Fluorescence emission spectra were recorded in a SPEX FluoroMax spectrofluorometer in SF-50 buffer (50 mM Tris–HCl pH 7.5, 50 mM NaCl, 1 mM DTT, 5 mM MgCl2) plus 10% glycerol. Fluorescein (FLU) and hexachlorofluorescein (HEX) emission was detected at 500–550 and 542–580 nm, with 494-nm and 538-nm excitation, respectively.
Transient kinetic measurements were carried out in a KinTek SF-2004 stopped-flow apparatus. Post-mix concentrations are stated in the text. Pi release measurements were performed in SF-150 buffer (SF-50 buffer plus NaCl to 150 mM). A Pi mop (150 µM 7-methylguanosine, 0.1 U/ml purine nucleoside phosphorylase) was present in all solutions. MDCC-PBP fluorescence was excited at 436 nm, and emission was followed through a 455-nm cutoff filter. Single-round translocation experiments were performed in the presence of heparin (see also Supplementary Data) in SF-50 buffer. 3′-(N-methylanthraniloyl)-2′-deoxy-ATP (mdATP) binding and release were measured in SF-50 buffer and mant fluorescence was detected through a 400-nm cutoff filter using 280-nm excitation, utilizing Förster resonance energy transfer (FRET) from BLMHM’s aromatic residues. Trp fluorescence was excited at 297 nM, and emission was detected through a 340-nm interference filter.
Steady-state ATPase activities of BLMHM (35 nM) were measured using a pyruvate kinase/lactate dehydrogenase (PK/LDH) coupled assay (14 U/ml PK, 20 U/ml LDH, 1 mM ATP, 1 mM phosphoenol pyruvate, 200 μM NADH) in SF-50 buffer plus 50 µg/ml BSA. Time courses of NADH absorbance (ε340 nm = 6220 M–1 cm–1) were followed in a Shimadzu UV-2101PC spectrophotometer. Oligonucleotide sequences are shown in Supplementary Data.
Data analysis, fitting and simulations were performed using OriginLab 7.5, KinTek SF-2004, and Gepasi v3.30 (www.gepasi.org). If not otherwise stated, reported uncertainties are standard errors of NLLS fits.
In this study, we used BLMHM (amino acids 642–1290), a fully active monomeric form of BLM that retains all activities and substrate specificities of the full-length enzyme (26). Fluorescence titration of 54-mer ssDNA molecules labeled either at the 3′-end or at the 5′-end with BLMHM showed very similar hyperbolic binding profiles and apparent dissociation constants (Supplementary Figure S1A), demonstrating that nucleotide-free BLMHM binds randomly and non-cooperatively along ssDNA with no end preference. To further assess DNA-binding properties, BLMHM was mixed with equimolar amounts or 4-fold excess of 5′-labeled oligo-dT substrates of different lengths. The dependence of fluorescence change amplitudes on oligo-dT length corroborated random binding (Supplementary Figure S1B).
ATP binding to BLMHM was directly monitored in stopped-flow experiments using mdATP, a fluorescent ATP analogue (Supplementary Figure S1C). The mdATP concentration dependence of the observed rate constants (kobs) of single-exponential binding transients indicated that ATP binding (step 1 in Figure 1B) is an apparently single-step, rapid and reversible process (Table 1). mdATP binding transients were practically uninfluenced by the presence of ssDNA, showing that ssDNA and ATP binding to BLMHM are uncoupled, similarly to that observed recently with forked dsDNA substrate (27) and that for DbpA RNA helicase (33).
The kinetics of release of Pi produced during ATP hydrolysis were followed using MDCC-PBP, a fluorescently-labeled Pi binding protein (34). When BLMHM plus dT54 was mixed with different concentrations of ATP in the stopped-flow, MDCC-PBP fluorescence transients consisted of an exponential burst followed by a linear steady-state phase (Figure 2A). The kobs of the exponential burst (67 ± 6 s–1, SD for n = 7) was independent of ATP concentration. This rate constant was limited by Pi binding to MDCC-PBP, in concert with earlier reports (34) and our MDCC-PBP Pi binding calibration transients (Supplementary Data, Supplementary Figure S2). The amplitude of the burst showed hyperbolic dependence on ATP concentration (Figure 2B). The released Pi/BLMHM molar ratio saturated at around unity, which indicates that the ATP hydrolysis and Pi release steps (k2 and k3 in Figure 1B, Table 1) are practically irreversible, and the rate-limiting step of the enzymatic cycle during translocation occurs after Pi release. Steady-state kcat (maximal steady-state turnover rate constant) and KATP (half-saturating ATP concentration) values were determined from slopes of the linear phase of MDCC-PBP fluorescence transients (Figure 2B inset). KATP (20 µM) was close to the ATP binding Kd (17 μM) calculated from mdATP binding kinetics (Figure 2B inset, Supplementary Figure S1C, Table 1), indicating that KATP is dictated by reversible ATP binding.
In another set of experiments, BLMHM was preincubated with different concentrations of dT54 and then rapidly mixed with ATP (at saturating concentration) in the stopped flow. In the absence of dT54, a low-amplitude Pi burst (Aburst = 0.05 mol Pi/mol BLMHM) was observed with a kobs similar to those in the presence of dT54 (Figure 2C). This behavior implies that in DNA-free BLMHM, ATP hydrolysis and Pi release occur rapidly but ATP hydrolysis is unfavorable with an apparent equilibrium constant of 0.05 (=Aburst/(1 – Aburst)). (Pi release is quasi-irreversible in the presence of MDCC-PBP.) Traces in the absence and presence of dT54 at different concentrations consisted of exponential burst and linear steady-state phases (Figure 2C). The kobs of the exponential burst was independent of dT54 concentration and was identical to that in the experiments of Figure 2A. The released Pi/BLMHM molar ratio calculated from the Pi burst amplitudes saturated at around unity (Figure 2D), again corresponding to rapid pre-steady-state ATP hydrolysis before the steady-state phase. The steady-state ATPase activity of BLMHM, calculated from the slope of the linear phase, was markedly activated by dT54 (Figure 2D inset, cf. Figure 2B inset).
The ssDNA length-dependence of the Pi release profile was determined by preincubating 0.25 µM BLMHM with 2 µM oligo-dT substrates of different length (dT18, dT36 and dT54), followed by rapid mixing with 500 µM ATP in the stopped flow. The amplitude of the rapid exponential burst of the MDCC-PBP fluorescence transients was practically identical in dT18, dT36 and dT54 (the released Pi/BLM molar ratio was 1.2 ± 0.5, SD for n = 3), again indicating pre-steady-state ATP hydrolysis (cf. Figure 2B and D).
We followed the transient kinetics of mdADP release from BLMHM in ‘chasing’ experiments in which we rapidly mixed a premixture of BLMHM, mdADP and varying concentrations of oligo-dT substrates of different length (dT18, dT36 and dT54) with excess unlabeled ATP in the stopped flow. In the absence of DNA, the dissociation of mdADP from BLMHM resulted in a biphasic fluorescence decrease with kobs values of 29 s–1 and 0.38 s–1, with the rapid phase having a fractional amplitude of 72% (Figure 3A). This result indicates that the BLMHM.ADP complex adopts two states in solution: the rapid phase corresponds to an ‘open’ state from which ADP can freely dissociate, whereas the slow phase indicates a ‘closed’ subpopulation that must first slowly convert to the open state to release ADP (step 4 in Figure 1B). This conclusion is in line with previous implications that the stimulatory effect of DNA on the BLMHM ATPase activity is brought about by an enhancement of the rate of the ADP dissociation process (27). Importantly, however, here we directly show that it is not the ADP dissociation step itself but a conformation change preceding ADP release that is rate limiting during the translocation cycle. In line with this, mdADP binding to apo-BLMHM resulted in biphasic transients; the presence of the slow phase (with a kobs ~0.4 s–1) confirmed the slow isomerization reaction from the closed to open state of BLMHM (Supplementary Figure S3). In the presence of increasing concentrations of oligo-dT, the kobs of the rapid and slow phases in the ‘chasing’ experiments increased to 270 ± 30 s–1 and 5.3 ± 0.6 s–1, respectively (Figure 3A). In the presence of DNA, mdADP binding was also biphasic with a slow phase kobs of 8.1 ± 0.3 s–1 [corresponding to (k4 + k–4) in Figure 1B], which again corroborates that the closed-open transition (represented in the slow phase) is the rate-limiting step during the steady-state translocation cycle (Figure 1B, Table 1). The rate constants of the isomerization phase were somewhat lower than the maximal DNA-activated ATPase activity of BLMHM in Pi release experiments (Figure 2D inset). This difference can possibly result from the presence of the fluorescent label in mdADP. [Besides several studies on cytoskeletal motor proteins (35–37), it was recently reported that mantADP binds with a >40-fold higher affinity to PcrA than unlabeled ADP (38).]
The presence of ssDNA had a marked effect on the magnitude of total fluorescence changes upon mdADP release. The dependence of this parameter on oligo-dT concentration turned out to be a highly useful indicator of the stoichiometry of BLMHM binding to oligo-dT substrates of different length (Figure 3B). In contrast to recent results (27), the BLMHM.ADP complex exhibited high-affinity oligo-dT binding profiles, from which we deduced BLMHM/oligo-dT molar ratios of 0.95 ± 0.17, 2.0 ± 0.2, and 3.8 ± 0.6 for dT18, dT36 and dT54, respectively (Figure 3B). The result for dT54, which is the least affected by DNA end-effects, indicates that a single BLMHM molecule occupies an ssDNA stretch of about 14 nt (parameter b in Figure 1A, Table 1). This binding site size was confirmed in other experiments (see below).
In the experiments of Figure 3, the kobs of mdADP fluorescence transients and the fractional amplitudes of the two phases were independent of mdADP concentration (10–100 µM) in all cases. [The rapid phase had a fractional amplitude of 79 ± 4% (SD for n = 10) at 10 µM mdADP and 71 ± 5% at 100 µM mdADP.] These results, together with those of Supplementary Figure S3, indicate that the experiments of Figure 3A truly report on the process of multistep mdADP dissociation instead of phenomena related to multiple binding sites or binding modes.
Heparin has been used as a protein trap to generate single-round conditions in which the helicase cannot rebind to the DNA track after dissociation (20,30,39). Heparin has no ATPase activity, and accelerates DNA-free BLMHM ATPase activity only 2.5-fold (Supplementary Figure S4). When BLMHM was preincubated with saturating concentrations of oligo-dT substrates of different length and then rapidly mixed in the stopped-flow with ATP plus heparin, the MDCC-PBP fluorescence transients showed a multiphasic Pi release profile (Figure 4A). Traces with substrates longer than the binding site size consisted of a rapid exponential burst and two distinct linear phases, while in those with dT12 the first, more rapid linear phase was lacking (Figure 4A, inset). The amplitude of the exponential burst was practically equivalent to the pre-steady-state hydrolysis of one ATP per enzyme molecule (Aburst = 1.1 ± 0.1 mol Pi/mol BLMHM) as previously seen in the absence of heparin (Figure 2), and its kobs (55 ± 12 s–1, SD for n = 12) was also similar to those obtained in the absence of heparin. This exponential phase therefore represents the first round of ATP hydrolysis by DNA-bound BLMHM. The ATP turnover rates calculated from the first linear phase were independent of oligo-dT length above 12 nt (27 ± 2 s–1, SD for n = 12). The amplitudes of this phase increased with oligo-dT length (Figure 4B), indicating that they represent ATP consumption during translocation of BLMHM along ssDNA. Based on our recently published analytical method (40), the amplitude data revealed that the presence of heparin decreased the processivity of BLMHM (Figure 4B; see also Supplementary Data). The processivity at zero heparin concentration was P = 0.98 (expressed as the probability of a translocation step), corresponding to a mean run length (P/(1 – P)) of 50 steps (Figure 4C, see also Supplementary Data) (40). The step size (s = 1.1 ± 0.1 nt traveled per ATP hydrolyzed, corresponding to an ATP coupling ratio of 0.87 ATP/nt) and binding site size (b = 12.4 ± 0.1 nt) were independent of heparin concentration (Figures 4B and 1A, Table 1). The binding site size was very similar to that determined in the experiments of Figure 3B. The ATP consumption rate during the second linear phase (0.14 ± 0.08 s–1, SD for n = 12) was similar to the steady-state ATPase rate of heparin-bound BLMHM (Supplementary Figure S4), indicating that this phase mainly corresponds to steady-state ATP hydrolysis by heparin-bound BLMHM after it has dissociated from DNA. In line with this, this value also matches the ATPase rate in the linear phase in experiments with dT12 where the lack of a more rapid linear phase indicated the lack of translocation due to the short length of DNA (Figure 4A).
To corroborate the translocation parameters described above, we also performed experiments monitoring Pi release during single-round translocation of BLMHM on m13 circular ssDNA (Supplementary Figure S5), as well as Trp fluorescence experiments monitoring the dissociation of BLMHM from ssDNA (Supplementary Figure S6).
The steady-state ATPase activity of DNA-free BLMHM, determined using a PK/LDH coupled assay (kcat = 0.081 ± 0.007 s–1, SD for n = 8, KATP = 26 ± 3 µM), was strongly activated by non-repetitive sequence oligonucleotides (kcat, DNA_54mer = 13 s–1), oligo-dT (dT5 – dT90) (Figures 2 and and5)5) and m13mp18 circular ssDNA (kcat, m13 = 13 s–1). From these data, it is obvious that oligo-dT activated the ATPase activity slightly more (within a factor of two) than non-repetitive sequence oligonucleotides of the same length and m13 phage ssDNA. Moreover, oligo-dT90 substrate with phosphorothioate (PTO) modification also stimulated the ATPase activity of BLMHM with similar degree as the unmodified dT90 (kcat, PTO-dT90 = 20 s−1) (Figure 5).
We titrated BLMHM with increasing oligo-dT concentrations to determine the DNA concentration necessary for half-maximal ATPase activation (KDNA) at saturating ATP concentration. KDNA steeply decreased with increasing oligo-dT length until the length exceeded the binding site size (b = 14 nt) (Figure 5A (inset); cf. Figures 3B and 4B). At longer lengths, the shallow decrease in KDNA was mostly dictated by the binding stoichiometry of BLMHM to oligo-dT substrates of different length (cf. Figure 3B). This was even more apparent when the KDNA data were converted into the molar excess of DNA binding sites over BLMHM (Figure 5A).
Importantly, the maximal ssDNA-activated ATPase turnover rates (kcat) showed a characteristic length-dependent profile (Figure 5B). The kcat values using longer oligo-dT substrates are in good agreement with those measured recently with full-length BLM (7) indicating that the monomeric form of BLM utilizes the same translocation mechanism as the full-length form. kcat stagnated in the length range of 7–12 nt (~5 s–1, corresponding to kend in Figure 1A, Table 1), then steeply increased to a maximal value of ~25 s–1 (confirming that the first linear phase in Figure 4A corresponds to the ATP consumption rate during translocation along ssDNA), already reaching quasi-saturation at a length of ~60 nt. This profile matches the predictions expected from the kinetic model of BLMHM translocation based on the presented experiments (Figure 5B legend, Table 1, ‘Discussion’ section and Supplementary Data).
The KDNA values of Figure 5A also support that, in kinetic experiments performed at a single DNA concentration (Figures 2A and B, 4, 5B, S1C, S3 and S5 and 6), the applied DNA concentration (typically micromolar) was quasi-saturating and high enough to rule out the possibility that the results were influenced by simultaneous binding of multiple BLMHM molecules to the same DNA molecule.
During an enzymatic cycle leading to translocation along ssDNA (occurring at a net turnover rate ktrans in Figure 1A), ATP binding is followed by hydrolysis and product release steps (Figure 1B, Table 1). Our results reveal the following salient features of this cycle: (i) ATP binding is rapid, reversible, and unaffected by ssDNA (Figure 2 and Supplementary Figure S1C). (ii) In ssDNA-bound BLMHM, ATP hydrolysis and Pi release are rapid, quasi-irreversible, and precede the rate-limiting step (Figures 2 and 4A). (iii) The rate-limiting step (probably a conformational transition) occurs between two ADP-bound BLMHM states (Figure 3A). This step is markedly accelerated by ssDNA (Figure 3A).
Processive translocation along DNA requires multiple ATP hydrolysis-coupled DNA-bound translocation steps. Figure 1A and Table 1 summarize the proposed translocation mechanism of BLMHM, exhibiting the following key features: (i) DNA binding occurs randomly along ssDNA (Supplementary Figure S1A and B). (ii) The binding site size (b) is about 14 nt (Figures 3B, 4A and B and 5). (This is not necessarily equivalent to the contact site size of the BLMHM–ssDNA complex; it is defined as the number of occluded nucleotides by a BLMHM molecule.) This binding site size is somewhat larger than that reported recently (27). However, contrary to the results presented here, in that study the binding site size was determined only from the fit to anisotropy titration data, without confirmatory independent results. (iii) During one ATP hydrolysis cycle (occurring at a net rate ktrans), BLMHM travels 1 nt along ssDNA (s) (Figure 4A and B). (iv) On reaching the 5′-end (where the ATPase cycle slows down to kend), BLMHM rapidly dissociates from ssDNA (koff, end) (Figure 5B). (v) The rate constant of BLMHM dissociation from ssDNA during translocation (koff, int) is relatively low, resulting in a moderate translocation processivity (with a mean of 50 nt traveled in a single run) (Figure 4C, Table 1).
BLM can unwind dsDNA regions shorter than 100 nt in the absence of binding partners (6). BLMHM was recently shown to unwind forked DNA substrates with low processivity (27), indicating that its unwinding processivity is somewhat lower than the translocation processivity determined in this study. However, the difference may at least partially result from the fact that the dependence of unwinding processivity on trap concentration was not determined (27). Interestingly, the unwinding step size of Escherichia coli RecQ helicase was 4 nt (41), a value that greatly differs from the unwinding step size of BLMHM (27) and the translocation step size reported here. However, a direct measurement of ATP consumption during unwinding is still lacking. The available data highlight the necessity of precise measurements of enzymatic parameters during dsDNA unwinding.
The kinetic mechanism of BLM translocation along ssDNA described here, the first one for a human helicase, displays several features that are different from those of other helicases. The key difference between the previously described translocation mechanisms of PcrA and UvrD (SF1) (30,42), RecG and Isw2 (SF2) (43,44) and that of BLM is that BLM performs DNA length-independent rapid pre-steady-state ATP hydrolysis before steady-state translocation (Figures 2 and 4A). This behavior is due to rapid, non-rate-limiting ATP binding, hydrolysis and Pi release (Table 1). In SF1 helicases, steady-state translocation was not preceded by a pre-steady-state burst in Pi production, indicating that the rate-limiting step precedes Pi release (30,42).
Another key property of the BLM mechanism is the low ATP coupling ratio (Figure 4B, Table 1). In line with PcrA and UvrD in which the hydrolysis of one ATP was observed per one translocated nucleotide on ssDNA (based on both solution measurements and on crystal structures), BLM also travels 1 nt per ATP hydrolysis. Although RecG was recently shown to exhibit a low ATP coupling ratio (1 ATP/3 nt), this indirect measurement was based on translocation rate during unwinding (44), which might significantly differ from that during ssDNA translocation.
A third important feature is that, after completing translocation, BLM does not perform multiple rounds of ‘futile’ ATP hydrolysis at the 5′ end. The characteristic ssDNA length-dependence of the ATPase activity (Figure 5B) indicates that BLM relatively rapidly dissociates from the 5′-end (at koff, end) where its ATPase cycling rate (kend) is lower than that during translocation (ktrans) (Figure 1A, Table 1). Although PcrA had been proposed to remain bound at the 5′-end and perform several slow enzymatic cycles before dissociation (very slow koff, end), which resulting in a ssDNA length-independent steady-state ATPase activity governed by kend (42), using longer DNA substrates PcrA also showed DNA length dependence of steady-state kcat (38), indicating a mechanism similar to the one proposed in the current study. On the other hand, the ssDNA length-independent ATPase activity of UvrD (22) results from a different set of features, namely that the ATPase rates during translocation and at the ssDNA end are practically identical (ktrans = 42 s–1, kend = 40 s–1) (20,30).
The translocation mechanisms of PcrA, UvrD and Rep (SF1) have revealed an inchworm mechanism in which nucleotide bases of ssDNA form hydrophobic stacking interactions with aromatic residues of the enzyme, and nucleotide-induced conformational changes displace the ssDNA strand along the DNA-binding cavity (19,28,45). In contrast, recent studies on SF2 helicases (hepatitis C virus NS3, archaeal Hel308) support a translocation mechanism based on protein-DNA backbone interactions (15,46,47). BLM contains the tryptophan (Trp803) that, based on its interaction with DNA nucleotide bases, has been identified as a key residue for the hydrophobic stacking-based mechanism and is conserved through SF1 and SF2 helicases (sequence alignments of the corresponding region of SF1 and SF2 helicases are shown in Supplementary Data, Table S1).
On the other hand, BLM also contains key structural elements proposed for a translocation mechanism based on enzyme-DNA backbone interactions. In existing atomic structures of DNA-bound viral NS3, Drosophila VASA and eIF4AIII (SF2) enzymes, two conserved threonines have been shown to play fundamental roles in helicase-DNA backbone interactions (47–51). One of these threonines is located in motif V, which is conserved among SF1 and SF2 enzymes. Structural and/or sequence similarities between motif V of NS3, Swi2/Snf2, RecQ and BLM suggest that this conserved threonine (Thr946 in BLM) may play a similar role in BLM translocation as in NS3. The recently published atomic structure of DNA-bound RecQL1 (PDB code 2WWY) also reveals that this threonine (Thr371 in RecQL1) is in direct contact with the backbone of the ssDNA strand spanning the surface of the RecA fold domains. The other conserved threonine is located in motif Ib of NS3, a region missing from RecQ helicases. Further structural similarities between NS3, Hel308, RecQ, and RecQL1 support the hypothesis that ssDNA binds to the large cleft formed by the two RecA fold domains and the RecQCt domain and that BLM residues may form contacts to the sugar-phosphate backbone of ssDNA.
Another possible mechanism for BLM translocation is a ‘mixed mechanism’, which was recently published for an SF1B helicase (52). In this case, both hydrophobic stacking and backbone interactions determine the translocation with 1 nt per one ATP hydrolysis (though in the 5′-3′ direction). This assumption is corroborated by the fact that in our experiments, PTO-modified oligo-dT90 activated BLMHM to a similar degree as the unmodified oligo and KDNA was also only slightly altered (Figure 5). This result indicates that BLM can translocate on DNA substrates with modified sugar phosphate backbone, but the translocation is somewhat slower.
The slow, rate-limiting conformational transition in ADP-bound BLMHM identified in this work may represent the opening of the nucleotide-binding cleft formed by the two RecA-folds of the helicase module, which could result in a physical step along ssDNA. By contrast, a rate-limiting Pi release step was proposed to drive the power stroke used for unwinding by YxiN and NS3 RNA helicases (53,54).
It was recently proposed that DNA weakens the affinity of BLMHM for ADP, thus accelerating the steady-state ATPase rate. Our present study points out that ADP release itself occurs rapidly, and the allosteric activation of the BLMHM ATPase by DNA is brought about specifically via the acceleration of the slow structural transition preceding ADP release, presumably by the interaction of DNA with the large cleft spanning the whole length of the RecA domains. Crystal structures and homology modeling of SF1 and SF2 helicases (19,28,47,55,56) support that this structural change may enable ADP dissociation and stepping along the DNA track in an inchworm-like manner.
The present general framework for unwinding activities discriminates between two catalytic strategies: a ‘passive’ one in which the helicase moves forward and unwinds the substrate by exploiting the thermal fraying of DNA base pairs, in contrast to ‘active’ unwinding whereby the enzyme facilitates the destabilization of the duplex ahead of it (15,57–59). Based on this model, enzymes using a passive mechanism unwind double-stranded nucleic acid segments at significantly reduced rates compared to that of translocation along single-stranded nucleic acids. In contrast, optimally active enzymes can separate the two strands at the same rate as the translocation rate. Importantly, the ssDNA translocation rate of BLMHM determined in this study is very similar to that reported for dsDNA unwinding by the same construct (27), which supports a fully active model for the dsDNA unwinding mechanism of BLM.
Supplementary Data are available at NAR Online.
Funding for open access charge: The Hungarian Scientific Research Fund (71915 to M.K.); and Norway Grants (78783 to M.K.).
Conflict of interest statement. None declared.
We thank Lumir Krejci for the BLM-coding cDNA clone and for discussions. M.K. is a Bolyai Fellow of the Hungarian Academy of Sciences.